Add Claude and OpenCode Cognitive Infrastructure with Knowledge Packs

[stephanchenette]

Summary

This PR adds comprehensive cognitive infrastructure setup playbooks for both Claude Code and OpenCode AI coding assistants, along with knowledge packs for extending agent capabilities.

What's Included

Claude Cognitive Infrastructure (Setup/Claude-Cognitive-Infrastructure/)

• Hook System with 10-tier security validation
• Memory System (three-tier: hot/warm/cold)
• Skills Framework with USE WHEN activation
• Complete 6-phase playbook (0_INITIALIZE through 5_VERIFY)
• Knowledge Pack: Codebase-Expert for RAG-powered code understanding

OpenCode Cognitive Infrastructure (Setup/OpenCode-Cognitive-Infrastructure/)

• Plugin System (TypeScript-based event handling)
• Memory System (three-tier architecture)
• Skills Framework (lowercase naming convention per OpenCode spec)
• Complete 6-phase playbook
• Knowledge Pack: codebase-expert for RAG-powered code understanding

Structure

Setup/
├── Claude-Cognitive-Infrastructure/
│   ├── 0_INITIALIZE.md ... 5_VERIFY.md
│   ├── README.md
│   ├── assets/
│   └── Knowledge-Packs/
│       └── Codebase-Expert/
└── OpenCode-Cognitive-Infrastructure/
    ├── 0_INITIALIZE.md ... 5_VERIFY.md
    ├── README.md
    └── Knowledge-Packs/
        └── codebase-expert/

Key Features

Feature	Claude	OpenCode
Config Directory	.claude/	.opencode/
Identity File	CLAUDE.md	AGENTS.md
Hooks/Plugins	Shell commands	TypeScript plugins
Skill Naming	TitleCase	lowercase-with-hyphens
Security	10-tier validator	10-tier validator

Test plan

• Run Claude playbook on test agent directory
• Run OpenCode playbook on test agent directory
• Verify all directories and files created correctly
• Test hook/plugin functionality
• Verify skill discovery works

🤖 Generated with Claude Code

[pedramamini]

Review Feedback

@stephanchenette - Great contribution! The playbooks are well-structured and the documentation is thorough. A few suggestions to consider before merging:

Key Items

1. Placeholder embedding implementation - The _Core/lib/embeddings.ts returns random values as a placeholder. The README should clarify that users need to configure an actual embedding provider (OpenAI, local models, etc.) for the Codebase Expert pack to function.

2. Hook registration mechanism - The RAG hook (rag-retrieval.ts) is written but there's no mechanism in Claude Code to actually execute these hooks. Could you clarify: is the intent that users manually integrate these, or is there a planned hook execution system?

3. Skill activation documentation - The SKILL.md frontmatter assumes a trigger-based activation system (triggers: [identity, who are you]). It would help to document:

   • How skills get activated (manual/automatic?)
   • What consumes the context_budget field
   • Whether this integrates with Claude Code's existing systems

4. Gitignore guidance - Should users commit .claude/context/memory/? These files may contain sensitive session data. Consider adding a note about this.

Minor Suggestions

• Add migration guidance for future version upgrades (VERSION file starts at 1.1.0)
• Consider adding a JSON schema for settings.json validation
• The manual "Generate embeddings" step in 4_IMPLEMENT.md needs an actual embedding provider to work

Overall

This is solid work that adds real value. The main gap is clarifying what's automated vs. what requires user configuration. Happy to approve once the documentation addresses the embedding provider requirement and skill/hook activation mechanisms.